Image-based solutions for gesture recognition or hand pose detections are becoming more sophisticated every year. Current solutions include using global color models to detect skin. An issue with this solution may include a high error rate. Other solutions for detecting hand poses use trained regression models. A drawback of trained regression models is the large amount of resources expended, time needed, or difficulty in training the models. Still other approaches include using skin information to directly detect hand poses. This approach may not work with all skin colors and may require significant resources or time.